Real world delivery of cancer precision medicine: The UCSF experience

E. Alejandro Sweet-Cordero, MD Benioff UCSF Professor of Children’s Health Head, UCSF Molecular Oncology Initiative

PMWC 2019 Mountain View, CA

e Precision Cancer Medicine

Advanced Cancers

Comprehensive Profiling

Molecular Tumor Board

Decision Support Systems

Tumor

Blood

Treatment Guidance

Individualized treatment, umbrella/basket trials

Resistance?

Better Predictions Improved Outcomes Less Toxicity

e UCSF500: Paired normal/tumor DNA next-generation sequencing assay

• CLIA-approved – available to patients seen across UCSF sites

e UCSF500 Technical specifications

• UCSF pediatric AND adult oncology community consulted for

inclusion of genes relevant to their diseases of interest

• Tumor DNA extracted from FFPE + normal DNA

• Exon tiling across 479 genes

• Selected intron tiling across 47 genes.

• Selected probes across genome to report copy number

• Average read depth 500X

• Variant pipeline in DNAnexus

• Report generated using Genome Oncology platform

e UCSF500 Panel Assay

• Current volume ~100 sequenced cases/month

• First case signed-out April of 2015

• >2000 cases reported to date

• 80% Adult, 20% Pediatric

• Data imported and visualized using a local cbioportal

instance

• Manuscript reporting results on initial 2000+ cases in preparation

• Held every two weeks

• Cases discussed by physician request (4-8 cases/session)

• Both UCSF and outside referrals (Kaiser, UC Davis, Valley Children’s Hospital, many others)

• Cases discussed prepared/discussed by referring clinician, molecular pathologist and oncologist with molecular expertise (Moasser, Collisson, Stieglitz, Klein, Sweet-Cordero)

• Sessions attended by core group of pediatric/adult clinicians with wide disease-specific expertise as well as geneticists, pharmacists and others.

• Written report provided to clinicians

• IRB approved Registry Trial

e UCSF Molecular Tumor Board

Molecular Tumor Board

Change in Diagnosis

Biomarker-guided treatment options

Risk Stratification

Cancer Genetics (germline findings)

• Biweekly meetings • A team molecular

pathologists, genetic counselors and clinicians with genomics expertise research variants/genes in report

• Clinical hx of patient,

UCSF500 results, and recommendations discussed at meeting

• Recommendation report

provided

UCSF500 Report

Utility of Test

CLINICAL OUTCOME TRACKING

e UCSFMolecular Tumor Board

Clinical Utility of UCSF500 Testing Uncover cancer risk traits Detection of therapeutic targets Firmly establish diagnosis

e The UCSF500 Report

e UCSF500 – Molecular Tumor Board - Patient Flow

1276 cases analyzed

174/438 (40%) with MTB treatment

recommendations

793 (62%) cases not discussed

62/174 (36%) therapeutic action taken (e.g. Rx)

95/174 (55%) no therapeutic action

17/174 (10%) lost to follow-up

300 (24%) No therapeutic recommendation 483 (38%) discussed at MTB

Updated to April, 2018

e Case Study #1: Change in Diagnosis leading to appropriate targeted therapy

e MTB Case Study: Change in Diagnosis

1/15: 20 year old man presented with left-sided hip pain and intermittent fevers. Clinical Diagnosis: Ewing sarcoma.

3/15: chemotherapy initiated.

5/15: no response to treatment, ongoing daily fevers, weight loss 50 lbs, bed ridden. 6/15: UCSF500: EWSR1-ATF1 fusion, indicative of histiocytoma, not Ewing sarcoma.

6/15: Specific treatment (anti-IL-6 ) initiated. Fevers resolved.

12/15: Tumor resected

7% of all UCSF500 analyses result in changes of the diagnosis

e Case Study #2: CNS tumors change in diagnosis leading to management change

• 4-year-old girl diagnosed as anaplastic medulloblastoma by referring provider

• Synaptophysin/neurofilament neg. and patchy OLIG2 staining not seen in medulloblastoma

• Internal tandem duplication within exon 15 of BCOR

• Amended diagnosis of CNS high-grade neuroepithelial tumor with BCOR alteration

• Because newly described, challenging to ascertain prognosis or identify best therapy; however, indicative of importance of correct identificationpreviously would have been inaccurately identified as MB and therefore would never have known to look for reasons why didn’t respond or why outcome different without knowing actual diagnosis was wrong->how do we know what we don’t know?

e Case Study #2: CNS tumors change in diagnosis leading to management change

Medulloblastoma to CNS high-grade neuroepithelial tumor with BCOR alteration

Highly cellular tumor with rosettes, scattered pseudorosettes, and areas of necrosis

MTB Case Study: Change in diagnosis

Courtesy of Dr. Cassie Kline

Medulloblastoma to CNS high-grade neuroepithelial tumor with BCOR alteration

MTB Case Study: Change in diagnosis

Courtesy of Dr. Cassie Kline

e Case Study #3: Targeted therapy

MTB Case Study: Targeted Therapy

A 12 y/o boy with metastatic chordoma

February 2016 April 2017

EZH2 Inhibition Tazemetostat

Rutkowski et al., JNS Ped., 19:531–537, 2017

e Case Study #4: recurrent metastatic osteosarcoma with multiple potential therapeutic targets

• 18 yo F with metastatic osteosarcoma

• Treated with standard of care chemotherapy (cisplatin/doxorubicin/methotrexate)

• 99% necrosis at resection (consistent with a very good response)

• Developed metastatic disease in the lungs. Metastatectomy performed as clinically indicated. Viable tumor sent for sequencing.

• Second line chemotherapy (ifosfamide and etoposide),

discontinued due to poor quality of life.

e Case Study #4: recurrent metastatic osteosarcoma with multiple potential therapeutic targets

UCSF500 on diagnostic biopsy

e Case Study #4: recurrent metastatic osteosarcoma with multiple potential therapeutic targets

Multiple genes identified, consistent with recent work indicating these are comon Copy number gained drivers in osteosarcoma (Sayles and Breese, Cancer Discovery 2019) Patient currently on treatment with CDK4 inhibitor

BRCA1/2, BRCAness PARPi, Platinum

MYC Dinaciclib*

PI3K BYL719*

HER2 Trastuzumab/Pertuzumab

IDH1,2 AG-120*, AG-188* FGFR2 BGJ398*

PDGFRA

SOS

BRAF

H3F3A

Everolimus, Temsirolimus

Dasatinib

Vemurafenib

Trametinib, Cobimetinib

Sorafenib

Peptide Vaccine* NF1

High TMB Pembrolizumab, Nivolumab SMARCB1 Tazemetostat*

Panobinostat

Cetuximab, Panitumumab

KRAS, RASA1

EGFR

TSC1/2, STK11

FLT3

ROS, ALK Crizotinib, Alectinib

Breast

Prostate

Endometrial/Ovarian

CNS

Lung

Gastro-esophageal

Colorectal

Pancreatic

Hepato-biliary

Neuroendocrine (GI)

RCC

Melanoma

JMML *Experimental Agent

e Matching Disease-Gene-Drug

e Cbioportal: Disease-specific data exploration

Summary 79 Diffuse Gliomas

e Cbioportal: Cross-disease gene centric analysis

e UCSF500 demographics

527, 78%

81, 12%

69, 10%

1, 0%

Pediatric Cases (n=678)

Somatic Only Somatic and Germline No Pathogenic Variants Detected Germline Only

1186, 84%

151, 11%

65, 5%

Adult Cases (n=1402)

Detected Variants in UCSF500 Cases e UCSF50: informative vs non-informative cases

90-95% of samples had an informative assay (likely pathogenic or pathogenic alteration somatic alteration AND/OR a germline alteration found).

e Most commonly altered somatic genes adult vs. pediatric

=not seen in adult

e Most commonly altered germline genes adult vs. pediatric

e Most common focal copy number alterations in UCSF500

e Long “Tail” of diagnosis in pediatric cancers

e Long “Tail” of fusion genes

APeX Clarity Caboodle RDB

Info Common

s (IC)

• Relational database • ~13,000

tables!

• Clinical Data Warehouse

• ~170 tables

Identified De-Identified

PHI

• Research Data Browser

• Flat file ‘datadump’

• Pt has10 digit surrogate identifier 1:1 with MRN

• Query-able database

• Structured eHR data

• 15 tables

”eLockBox”

(2) Pt MRN 10-digit RDB/IC ID

Mappings

MOI Database

• MRNs not stored!!

• Use RDB/IC ID for pulling patient-level data from IC

eLockBox: T. Goldstein

(1) 10-digit random RDB ID (de-ID) (2) Genetic alterations + test metadata

e Annotating genomic data using the EHR

Clinical Adoption Interpretation of

genomic results &

Clinical Care Recommendations

Capturing clinical outcomes

Knowledge sharing

Engage Insurance Providers

Clinical Trials

New Drug Development

Easily accessible, affordable & clinically-

impactful genetic testing

More actionable biomarkers

Engage Pharma

• Multi-institutional platform (e.g. GENIE)

• Peer reviewed publications

New Drug Approvals

e Precision Oncology Flywheel

e Does precision medicine lead to improved outcomes?

Days

P=0.007663

Kaplan-Meier Survival

Recommended treatment received?

Courtesy Jennifer Grabowski

Overview

International pancancer registry built through data sharing

Driven by openness, transparency, and inclusion

GOAL: improve clinical decision making • Linking clinical genotype to clinical outcomes

8 19

Eight founding participants, now 19 • North America & Europe • Plans for future expansion

Sponsored research Collaborative projects

e Project GENIE

e Project GENIE: participants

e Key challenges

• Building a knowledge base base on: • external data • analysis of internal data and prior experience

• Linking to clinical trials

• Clinical annotation of genomic results by data

extraction from EMR.

• Determing clinical utility and better definition of real-world barriers to utility

e

11/9/15 37

Molecular Oncology Unit

Summary

• The UCSF500 assay can detect a wide range of clinically relevant alterations that can modify therapy, refine diagnosis and identify inherited cancer predispositions.

• UCSF500 is unique in its coverage of both adult and pediatric alterations and paired germline/somatic analysis.

• Emerging opportunities for research in defining best practices to maximize utility of genomic testing.

• Data sharing through GENIE will increase opportunities for real world analysis of cancer precision medicine.

Acknowledgements

Michelle Turski Ted Goldstein Nhani Tran Jessica Van Ziffle Carlos Espinoza Courtney Onodera J.P. Grenert David Solomon

Cassie Kline Elliot Stieglitz Mark Moasser Eric Collisson Boris Bastian Iwei Yeh Jennifer Grabowski Amy Blanco

Molecular Oncology Initiative/CGL